Non-binding is not the same as inconsequential. The Trump administration’s National Policy Framework for Artificial Intelligence, released March 20, 2026, is a set of recommendations to Congress, it mandates nothing today. But it names a direction, identifies priorities, and proposes a legislative architecture that would reshape the compliance landscape if Congress moves on it.
The framework organizes its recommendations around seven pillars: child protection, AI infrastructure and small business support, intellectual property, censorship and free speech, enabling innovation, workforce preparation, and preemption of state AI laws. That last pillar is the one compliance teams need to track most closely.
The preemption language is specific. According to Morrison Foerster’s analysis, the framework calls on Congress to preempt state AI laws that “impose undue burdens,” while carving out state authority in areas including traditional police powers, child protection, and fraud prevention. What qualifies as an “undue burden” is undefined in the framework, that definition would be shaped by whatever legislation Congress ultimately produces, if any.
Timing is the complicating factor. One week after the White House released this framework, New York finalized the RAISE Act chapter amendment, a state AI law with a January 1, 2027 compliance deadline. The two documents point in opposite directions: one states pushing forward with enforceable frontier AI obligations, the other federal administration calling on Congress to clear those state obligations away. Compliance teams now face a landscape where building toward RAISE Act compliance and tracking a federal preemption push are happening simultaneously.
The framework builds on a December 2025 Executive Order that directed the administration to develop legislative recommendations and established an AI Litigation Task Force. That order is confirmed background context; the March 20 framework is the output. The intellectual property pillar has drawn separate analysis, the framework has been interpreted by analysts as endorsing the position that AI training on copyrighted material doesn’t infringe copyright law, though it defers resolution to the courts. For in-depth coverage of the federal AI copyright picture, see the hub’s prior reporting on competing White House and Senate copyright approaches and the three federal copyright framework analysis.
What to watch: Congressional action is the only mechanism that converts this framework into binding law. Track committee assignments, companion legislation, and whether any preemption language attaches to broader AI or technology bills. The “undue burdens” carve-out language is where the lobbying fight will concentrate, state legislators, civil liberties groups, and consumer protection advocates will contest any definition that sweeps too broadly. The DFS rulemaking timeline for the RAISE Act also becomes a preemption pressure point: the more New York’s rules are developed and in place, the harder a blanket preemption becomes to justify politically.
The framework’s significance is as a signal, not a statute. The administration has put federal preemption on the agenda. Whether Congress delivers is a different, longer question.